Evolving Mobile Robots in Simulated and Real Environments

By O. Miglino, H. Lund, Stefano Nolfi. 1995

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Deep learning in neural networks: An overview (2014, neutral): “as already mentioned in Sec. 1. Many RNN evolvers have been proposed (e.g., Miller et al., 1989; Wieland, 1991; Cliff et al., 1993; Yao, 1993; Nolﬁ et al., 1994a; Sims, 1994; Yamauchi and Beer, 1994; Miglino et al., 1995; Moriarty, 1997; Pasemann et al., 1999; Juang, 2004; Whiteson, 2012). One parti”
On Learning to Think: Algorithmic Information Theory for Novel Combinations of Reinforcement Learning Controllers and Recurrent Neural World Models (2015, methodology): “1 RL through Direct and Indirect Search in RNN Program Space It is possible to train small RNNs with a few 100 or 1000 weights using evolutionary algorithms [200, 255, 105, 56, 68] to search the space of NN weights [165, 307, 44, 321, 180, 259, 320, 164, 173, 69, 71, 187, 121, 313, 66, 270, 269, 305”
Crossing the reality gap in evolutionary robotics by promoting transferable controllers (2010, background): “The robustness of the behaviors can also be obtained by evaluating the solutions in diﬀerent simulated environments as in [21].”
Training Recurrent Networks by Evolino (2007, methodology): “more efficient and principled way is to search the space of RNN weight matrices [20,21,26,45,53,54] using evolutionary algorithms [1 3,31,41].”
Exploring the T-Maze: Evolving Learning-Like Robot Behaviors Using CTRNNs (2003, background): “1(a)) based on sensor sampling [6] and adding 5% uniform noise to the sampled values.”
Evolutionary Robotics and the Radical Envelope-of-Noise Hypothesis (1997, background): “Several experimenters, including Jakobi, Husbands, and Harvey (1995), Beer and Gallagher (1992), and Miglino, Lund, and Nolfi (1995) have shown that it is possible to evolve in simulation control architectures for a real robot.”
Leveraging multiple simulators for crossing the reality gap (2012, unknown): “…comparison of various simple test cases and found significant discrepancies between the performance of each simulator, primarily due to the design choices implemented in the dynamic simulation packages For simulating an underwater robot the fluid dynamics can be simulated with a number of…”
The Transferability Approach: Crossing the Reality Gap in Evolutionary Robotics (2013, background): “Similar robustness issues have been investigated in [40], also with Khepera mobile robots: by only choosing the most realistic amount of noise to add on the simulated sensor values, the transfer from simulation to reality did not lead to any performance loss.”
Evolution of Embodied Intelligence (2003, background): “Also adding some noise to the sensors and actuator while simulating the robot may help bridge the gap to reality [23] by avoiding that the controller over-specialises to the simulation.”
Levels of dynamics and adaptive behavior in evolutionary neural controllers (2002, methodology): “Given the simple experimental settings, the experiment is carried out in a realistic simulation by sampling infrared sensor activations (Miglino et al., 1995), computing geometric projections for the linear camera inputs, and adding 5% uniform noise to every value.”
Evolutionary design and behavior analysis of neuromodulatory neural networks for mobile robots control (2007, background): “To tackle the problem, many researchers have proposed to use techniques such as fitness-averaging [3] and/or adding noise [4,5] in the evolutionary process.”
Self-Organized Coordinated Motion in Groups of Physically Connected Robots (2007, background): “This setup implies that the s-bot’s wheels slip if motion is blocked by obstacles or by other connected s-bots.”
Evolution of Plastic Control Networks (2001, background): “Researchers have suggested solutions that under some conditions ensure a smooth transfer, such as sampling sensors and motors and storing values with added noise in lookup tables ( Miglino et al., 1996 ) or including in the simulation only those aspects of the interactions between the robot and the ”
Evolutionary robotics: coping with environmental change (2000, methodology): “The experiments have been carried out in simulations sampling sensor activation and adding 5% uniform noise to these values (Miglino et al., 1996).”
Gait optimization for roombots modular robots — Matching simulation and reality (2013, methodology): “RH measures the similarity of different trials of the same gait tested on the hardware by taking the pairwise similarity ratio SR between different trials (k,n ∈ [1,2,3]).”
Reinforcement Learning for UAV Attitude Control (2018, background): “nown as the reality gap, transferring from simulation to the real-world has been researched extensively as being problematic without taking additional steps to increase realism in the simulator [26], [3]. The vast majority of prior work has focused on performance of navigation and guidance. There is”